Network and Systems Laboratory nslab.ee.ntu.edu.tw Kaisen Lin, Aman Kansal, Dimitrios Lymberopoulos, and Feng Zhao Archiang
Network and Systems Laboratory nslab.ee.ntu.edu.tw Outline Introduction System Overview System Design Experiments and Evaluation Conclusions
Network and Systems Laboratory nslab.ee.ntu.edu.tw Introduction Popular mobile localization resource GPS, WiFi, Cell-tower ID, Bluetooth Continuous and ubiquitous location access aren’t available due to energy constraint Using multiple location sensors simultaneously to make up for this variability in accuracy would further increase energy use. Tradeoff between energy and accuracy
Network and Systems Laboratory nslab.ee.ntu.edu.tw Goal: a system that automatically manages location sensor availability, accuracy, and energy. GPS, WiFi, Cell-tower ID, Bluetooth Open sky view locations, availability and accuracy. Static and mobile Example Pizza stores in Portland Shopping Finding friends a - Loc
Network and Systems Laboratory nslab.ee.ntu.edu.tw System Overview Bayesian estimation Combine the sensor data and predicted location to provide a ML estimation Discretization A-Loc uses a 10m step size for space discretization. A-Loc uses time granularity of 1 minute Training
Network and Systems Laboratory nslab.ee.ntu.edu.tw System Design GPS, WiFi, Bluetooth, and cell-tower on Android G1 and AT&T Tilt phones Accuracy Models
Network and Systems Laboratory nslab.ee.ntu.edu.tw Energy Models
Network and Systems Laboratory nslab.ee.ntu.edu.tw Selection Algorithm The goal of the selection algorithm is to determine the most energy efficient sensor to be used, such that the required location accuracy can be achieved. This algorithm also maintains an estimate of the user’s location that is based on a prediction of user movements. Use Hidden Markov Model (HMM)
Network and Systems Laboratory nslab.ee.ntu.edu.tw Experiments and Evaluation Prototype Implementation Android’s LocationManager API Application Accuracy Requirement
Network and Systems Laboratory nslab.ee.ntu.edu.tw System Performance Accuracy requirement A-Loc compares with Static Least energy consumption Periodic Perfect Models Best resolution
Network and Systems Laboratory nslab.ee.ntu.edu.tw In San Diego
Network and Systems Laboratory nslab.ee.ntu.edu.tw In Portland
Network and Systems Laboratory nslab.ee.ntu.edu.tw Conclusions The authors present a-Loc system that can automatically tunne the location energy and accuracy trade-off by continually adapting to the dynamic location sensor characteristics and application needs. A-Loc provides significant energy savings that go beyond existing techniques.